Skip to content

Text classification by neural network models uses a dataset collected from Uzbek news sites.

Notifications You must be signed in to change notification settings

UlugbekSalaev/Text_Classification

Repository files navigation

Text classification dataset and analysis for Uzbek language

Multi-label text classification dataset for Uzbek language and some sourcode for analysis. This repository contains the code and dataset used for text classification analysis for the Uzbek language. The dataset consists text data from 9 Uzbek news websites and press portals that included news articles and press releases. These websites were selected to cover various categories such as politics, sports, entertainment, technology, and others. In total, we collected 512,750 articles with over 120 million words accross 15 distinct categories, which provides a large and diverse corpus for text classification. It is worth noting that all the text in the corpus is written in the Latin script.

Categories (with the name in Uzbek):

  • Local (Mahalliy)
  • World (Dunyo)
  • Sport (Sport)
  • Society (Jamiyat)
  • Law (Qonunchilik)
  • Tech (Texnologiya)
  • Culture (Madaniyat)
  • Politics (Siyosat)
  • Economics (Iqtisodiyot)
  • Auto (Avto)
  • Health (Salomatlik)
  • Crime (Jinoyat)
  • Photo (Foto)
  • Women (Ayollar)
  • Culinary (Pazandachilik)

Dataset

The dataset files can be accessed and downloaded from https://doi.org/10.5281/zenodo.7677431

Code

The code for text classification analysis is provided in this repository. We used Python programming and scikit-learn libraries for preprocessing and classifying the texts.

Results

Based on the model performance results, it can be concluded that the logistic regression models work best when both the word level and character level n-grams are considered (by concatenating their TF-IDF matrices). Neural network models, such as RNN and CNN , perform better than rule-based models, and their performance is enhanced by adding specific knowledge of the language, such as pretrained word-embedding vectors. Among the transformer-based models, the monolingual BERTbek model achieved the highest performance with an F1-score of 85.2%, compared to its multilingual counterpart (with 83.4% F1-score). The results of our experiments demonstrate the effectiveness of deep learning models for text classification in the Uzbek language and provide a strong foundation for further research in this area.

Citation

If you use this or paper in your research, please cite the following paper:

Kuriyozov Elmurod, Ulugbek Salaev, Sanatbek Matlatipov, & Gayrat Matlatipov. (2023). Text classification dataset and analysis for Uzbek language. 10th Language and Technology Conference: Human Language Technologies as a Challenge for Computer Science and Linguistics (LTC'23), Poznań. Poland.

Cite

@proceedings{kuriyozov_elmurod_2023_7677431,
  title        = {{Text classification dataset and analysis for Uzbek 
                   language}},
  year         = 2023,
  publisher    = {Zenodo},
  month        = feb,
  doi          = {10.5281/zenodo.7677431},
  url          = {https://doi.org/10.5281/zenodo.7677431}
}

Contact

For any questions or issues related to the dataset or code, please contact [elmurod1202@urdu.uz, ulugbek.salaev@urdu.uz].

About

Text classification by neural network models uses a dataset collected from Uzbek news sites.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages